UMG v. Udio

Case 1:24-cv-04777 Document 1 Filed 06/24/24 Page 1 of 41
UNITED STATES DISTRICT COURT

FOR THE SOUTHERN DISTRICT OF NEW YORK
UMG RECORDINGS, INC., CAPITOL Case No.: 24-04777

RECORDS, LLC, SONY MUSIC
COMPLAINT
ENTERTAINMENT, ARISTA MUSIC,
ARISTA RECORDS LLC, ATLANTIC
RECORDING CORPORATION, RHINO DEMAND FOR JURY TRIAL
ENTERTAINMENT COMPANY, WARNER
MUSIC INC., WARNER MUSIC
INTERNATIONAL SERVICES LIMITED,
WARNER RECORDS INC., WARNER
RECORDS LLC, and WARNER
RECORDS/SIRE VENTURES LLC,
Plaintiffs,
v.
UNCHARTED LABS, INC., d/b/a Udio.com,

and JOHN DOES 1-10,
Defendant.
Plaintiffs UMG Recordings, Inc. (“UMG”) and Capitol Records, LLC (“Capitol,” and
collectively with UMG, “Universal”); Sony Music Entertainment (“SME”), Arista Music, and
Arista Records LLC (“Arista Records,” and collectively with Arista Music and SME, “Sony”);
Atlantic Recording Corporation (“Atlantic”), Rhino Entertainment Company (“Rhino”), Warner
Music Inc. (“WMI”), Warner Music International Services Limited (“WMISL”), Warner Records
Inc., Warner Records LLC, and Warner Records/SIRE Ventures LLC (“WR/SIRE,” and
collectively with Atlantic, Rhino, WMI, WMISL, Warner Records Inc., and Warner Records LLC,
“Warner,” and together with Universal and Sony, “Plaintiffs”), by and through their undersigned
counsel, file this Complaint against Uncharted Labs, Inc. (“Udio”) and allege as follows:
NATURE OF THE ACTION
1. From the invention of the phonograph record, through the eras of vinyl, cassette
tapes, CDs, and now streaming and social media, the recorded music industry has been at the
forefront of technological advancement. Artificial intelligence (“AI”) and machine learning are
the next frontier of technological development, poised to push boundaries and expand commercial
opportunity. But with AI’s enormous capabilities comes an equally enormous potential for abuse,
making it critical that AI technology be implemented responsibly, ethically, and legally.
2. Most fundamentally, AI companies, like all other enterprises, must abide by the
laws that protect human creativity and ingenuity. There is nothing that exempts AI technology
from copyright law or that excuses AI companies from playing by the rules. This lawsuit seeks to
enforce these basic principles.
3. Perhaps more so than with many other technologies, there is both promise and peril
with AI. As more powerful and sophisticated AI tools emerge, the ability for AI to weave itself
into the processes of music creation, production, and distribution grows. If developed with the
permission and participation of copyright owners, generative AI tools will be able to assist humans
in creating and producing new and innovative music. But if developed irresponsibly, without
regard for fundamental copyright protections, those same tools threaten enduring and irreparable
harm to recording artists, record labels, and the music industry, inevitably reducing the quality of
new music available to consumers and diminishing our shared culture.
4. This case concerns a generative AI service, which allows users to generate digital
music files that sound like genuine human sound recordings in response to basic inputs. The
capacity for a generative AI service to produce convincing imitations of genuine sound recordings
starts with copying a vast range of sound recordings. When those who develop such a service steal
copyrighted sound recordings, the service’s synthetic musical outputs could saturate the market
with machine-generated content that will directly compete with, cheapen, and ultimately drown
out the genuine sound recordings on which the service is built.
5. Foundational principles of copyright law dictate that copying protected sound
recordings for the purpose of developing an AI product requires permission from rightsholders.
Otherwise, such AI offerings will erode the value of the artistic works that comprise the essential
-2-
raw materials that allow them to function in the first place. If left unmoored from existing and
longstanding legal constraints, such products could supplant, rather than support, genuine human
creativity.
6. Plaintiffs are record companies or recorded music businesses that, together, own or
exclusively control copyrights in a great majority of the most commercially valuable sound
recordings in the world. They have developed their enviable catalogs by discovering, developing,
and promoting human recording artists, whose artistic contributions are the bedrock of the
recorded music industry and the music we listen to today. These artists range from promising
newcomers to the most famous musicians and performers in the world to myriad other artists who
may not fill stadiums but who nevertheless shape culture. Plaintiffs have a track record of
embracing innovation and have entered into voluntary free-market licensing deals that authorize
the use of their protected sound recordings in emerging technologies. Such deals include full-
catalog licenses with streaming music services and user-generated content platforms, and other
licenses with innovative businesses associated with social media, fitness, gaming, the metaverse,
and more.
7. Defendant Uncharted Labs, Inc. is the company behind Udio, a generative AI
service that creates digital music files within seconds of receiving a user’s prompts. Building and
operating a service like Udio’s requires at the outset copying and ingesting massive amounts of
data to “train” a software “model” to generate outputs. For Udio specifically, this process involved
copying decades worth of the world’s most popular sound recordings and then ingesting those
copies into Udio’s AI model so it can generate outputs that imitate the qualities of genuine human
sound recordings. Udio charges many of its users monthly fees to use its product and produce
digital music files, which are designed to entertain, evoke emotion, and stoke passion just like the
genuine sound recordings Udio copied.
8. Given that the foundation of its business has been to exploit copyrighted sound
recordings without permission, Udio has been deliberately evasive about what exactly it has
copied. This is unsurprising. After all, to answer that question honestly would be to admit willful
-3-
copyright infringement on an almost unimaginable scale. Udio’s executives instead speak publicly
in exceedingly general terms. For example, they have said Udio’s service is trained on “a large
amount of publicly-available and high-quality music” that is “obtained from the internet” and the
“best quality music that’s out there.” 1
9. Of course, it is obvious what Udio’s service is trained on. Udio copied Plaintiffs’
copyrighted sound recordings en masse and ingested them into its AI model. Udio’s product can
only work the way it does by copying vast quantities of sound recordings from artists across every
genre, style, and era. The copyrights in many of those sound recordings—the recordings Udio’s
executives have described as the “best quality music” that can be obtained—are owned or
exclusively controlled by Plaintiffs. In other words, if Udio had taken efforts to avoid copying
Plaintiffs’ sound recordings and ingesting them into its AI model, Udio’s service would not be
able to reproduce the convincing imitations of such a vast range of human musical expression at
the quality that Udio touts. Udio’s service trains on the expressive features of these copyrighted
sound recordings for the ultimate purpose of poaching the listeners, fans, and potential licensees
of the sound recordings it copied.
10. If there were any doubt regarding Udio’s unauthorized copying, Udio dispelled it
by effectively conceding in pre-litigation correspondence that it copied Plaintiffs’ copyrighted
sound recordings. When Plaintiffs directly accused Udio of copying Plaintiffs’ sound recordings
to train its model, Udio did not deny or proffer any facts to undermine those allegations. It would
have been simple for Udio to say that it used other, legally acquired recordings, if that were the
case. Instead, Udio deflected and disingenuously asserted that its training data is “competitively
sensitive” and constitutes “trade secrets”—despite being based on “publicly available” music “out
there” for music fans. Udio also claimed that its large-scale copying of sound recordings is “fair
use,” which was telling because fair use only arises as a defense to an otherwise unauthorized use
of a copyrighted work.
1
Sharon Goldman, AI Music May Be Having a Moment, But Human Songwriters Would Like a Word, Fortune (May
17, 2024), https://fortune.com/2024/05/17/ai-music-training-scraped/.
-4-
11. Plaintiffs could have proceeded with this action based solely on eliciting that
reasonable inference of copying. Nevertheless, Plaintiffs’ claims are based on much more. In
particular, Plaintiffs tested Udio’s product and generated outputs using a series of prompts that
pinpoint a particular sound recording by referencing specific subject matter, genre, artist,
instruments, vocal style, and the like. Udio’s service repeatedly generated outputs that closely
matched the targeted copyrighted sound recording, which means that Udio copied those
copyrighted sound recordings to include in its training data. In addition, the public has observed
(and Plaintiffs have confirmed) that even less targeted prompts can cause Udio’s product to
generate outputs that resemble specific recording artists and specific copyrighted sound
recordings. Such outputs are clear evidence that Udio trained its model on Plaintiffs’ copyrighted
sound recordings.
12. Udio is not exempt from the copyright laws that protect human authorship. Like
any other market participant, Udio cannot reproduce copyrighted works for a commercial purpose
without permission. Heedless of this basic principle, Udio’s unauthorized copying erodes the
value and integrity of Plaintiffs’ copyrighted sound recordings with rapid and devastating impact.
Udio’s service generates music with such speed and scale that it risks overrunning the market with
AI-generated music and generally devaluing and substituting for human-created work. Despite
only having launched approximately two months ago, Udio’s product already makes a reported 10
music files per second, which amounts to 864,000 files per day, or just over 6,000,000 files per
week. These digital music files have been released to the public—some already finding their way
onto the major streaming services—and compete with the copyrighted sound recordings that
enabled their creation; yet Udio sought no permission from and gives no credit or compensation
to the human artists or other rightsholders whose works fueled their creation.
13. Udio also profits substantially from its infringement of Plaintiffs’ copyrighted
sound recordings. Udio has raised millions of dollars in funding, touts a roster of high-profile
backers, and has monetized its service, charging users up to $30 per month for its highest
subscription tier. None of that would be possible without the vast troves of copyrighted sound
-5-
recordings that Udio copied to train its AI model.
14. Udio cannot avoid liability for its willful copyright infringement by claiming fair
use. The doctrine of fair use promotes human expression by permitting the unlicensed use of
copyrighted works in certain, limited circumstances, but Udio offers imitative machine-generated
music—not human creativity or expression. Moreover, the Copyright Act enumerates four factors
to assess whether an unauthorized use is fair, none of which favors Udio’s product. These factors
are: (1) the purpose and character of the use; (2) the nature of the copyrighted work; (3) the amount
and substantiality of the portion used in relation to the copyrighted work as a whole; and (4) the
effect of the use upon the potential market for or value of the copyrighted work. In these
circumstances, the purpose of Udio’s use of Plaintiffs’ copyrighted sound recordings is
quintessentially commercial and creates directly competitive digital music files that serve the same
purpose as the recorded music Plaintiffs create and substitute for genuine recordings by humans;
Udio copies the key expressive features of Plaintiffs’ copyrighted sound recordings; those
copyrighted sound recordings are at the core of copyright protection; and Udio’s infringement
undermines both existing and potential commercial markets for selling, licensing, and distributing
sound recordings. If left unchecked, Udio risks upending whole segments of the legitimate music
industry.
15. At its core, this case is about ensuring that copyright continues to incentivize human
invention and imagination, as it has for centuries. Achieving this end does not require stunting
technological innovation, but it does require that Udio adhere to copyright law and respect the
creators whose works allow it to function in the first place.
16. Plaintiffs bring this action seeking an injunction and damages commensurate with
the scope of Udio’s massive and ongoing infringement.
THE PARTIES
17. Plaintiffs comprise the world’s foremost record companies and recorded music
businesses, engaged in the business of producing, manufacturing, distributing, selling, licensing,
and otherwise commercializing sound recordings in the United States and the world through
-6-
various media. Plaintiffs have made substantial investments in the development and promotion of
some of the most prolific and well-known recording artists in the world. Plaintiffs’ investments
extend further to lesser-known artists as well, with an eye toward sustaining the music industry
and discovering and supporting new generations of recording artists across all genres and styles.
Pursuant to their relationships with artists, Plaintiffs own or exercise exclusive control over rights
in millions of sound recordings of enormous cultural significance, artistic merit, and economic
value.
18. Plaintiff UMG Recordings, Inc. is a Delaware corporation with its principal place
of business at 2220 Colorado Avenue, Santa Monica, California 90404. UMG owns or exercises
exclusive control over the copyrights for the sound recordings within its catalog.
19. Plaintiff Capitol Records, LLC is a Delaware limited liability company with its
principal place of business at 2220 Colorado Avenue, Santa Monica, California 90404. Capitol
owns or exercises exclusive control over the copyrights for the sound recordings within its catalog.
A non-exhaustive list of specific sound recordings owned or exclusively controlled by Universal
that Udio has infringed is attached as Exhibit A (the “Universal Works”).
20. Plaintiff Sony Music Entertainment is a Delaware general partnership, the partners
of which are citizens of New York and Delaware. SME’s headquarters and principal place of
business are located at 25 Madison Avenue, New York, New York 10010. SME owns or exercises
exclusive control over the copyrights for the sound recordings within its catalog.
21. Plaintiff Arista Music is a New York partnership with its principal place of business
at 25 Madison Avenue, New York, New York 10010. Arista Music is a subsidiary of SME. Arista
Music owns or exercises exclusive control over the copyrights for the sound recordings within its
catalog.
22. Plaintiff Arista Records LLC is a Delaware Limited Liability Company with its
principal place of business at 25 Madison Avenue, New York, New York 10010. Arista Records
is a subsidiary of SME. Arista Records owns or exercises exclusive control over the copyrights
for the sound recordings within its catalog. A non-exhaustive list of specific sound recordings
-7-
owned or exclusively controlled by Sony that Udio has infringed is attached as Exhibit A (the
“Sony Works”).
23. Plaintiff Atlantic Recording Corporation is a Delaware corporation with its
principal place of business at 1633 Broadway, New York, New York 10019. Atlantic owns or
exercises exclusive control over the copyrights for the sound recordings within its catalog.
24. Plaintiff Rhino Entertainment Company is a Delaware corporation with its principal
place of business at 777 S. Santa Fe Avenue, Los Angeles, California 90021. Rhino owns or
exercises exclusive control over the copyrights for the sound recordings within its catalog.
25. Plaintiff Warner Music Inc. is a Delaware corporation with its principal place of
business at 1633 Broadway, New York, New York 10019. WMI owns or exercises exclusive
control over the copyrights for the sound recordings within its catalog.
26. Plaintiff Warner Music International Services Limited is a limited liability
company organized and existing under the laws of England and Wales with its principal place of
business at 27 Wrights Lane, London, England. WMISL owns or exercises exclusive control over
the copyrights for the sound recordings within its catalog.
27. Plaintiff Warner Records Inc. is a Delaware corporation with its principal place of
business at 777 S. Santa Fe Avenue, Los Angeles, California 90021. Warner Records Inc. owns
or exercises exclusive control over the copyrights for the sound recordings within its catalog.
28. Plaintiff Warner Records LLC is a Delaware corporation with its principal place of
business at 777 S. Santa Fe Avenue, Los Angeles, California 90021. Warner Records LLC owns
or exercises exclusive control over the copyrights for the sound recordings within its catalog.
29. Plaintiff Warner Records/SIRE Ventures LLC is a Delaware limited liability
company with its principal place of business at 1633 Broadway, New York, New York 10019.
WR/SIRE owns or exercises exclusive control over the copyrights for the sound recordings within
its catalog. A non-exhaustive list of specific sound recordings owned or exclusively controlled by
Warner that Udio has infringed is attached as Exhibit A (the “Warner Works”).
30. A non-exhaustive, illustrative sampling of Plaintiffs’ federally copyrighted sound
-8-
recordings that Udio has illegally reproduced is attached hereto as Exhibit A. Plaintiffs currently
commercially exploit, and at all relevant times have commercially exploited, all the sound
recordings listed in Exhibit A. Plaintiffs intend to amend the Complaint at an appropriate time to
provide an expanded list of works that Udio has infringed.
31. Defendant Uncharted Labs, Inc. d/b/a Udio is a Delaware corporation with its
principal place of business at 750 Lexington Avenue, Floor 9, New York, New York 10022.
32. Defendants John Does 1-10 are unknown parties who directly copied Plaintiffs’
federally copyrighted sound recordings, worked with Udio to copy Plaintiffs’ federally
copyrighted sound recordings, or have knowledge of Udio’s direct infringement of Plaintiffs’
federally copyrighted sound recordings and intentionally induced and materially contributed to the
infringement by assisting Udio’s compiling, scraping, and/or copying of the copyrighted sound
recordings for Udio’s training data, intentionally promoted or encouraged Udios’s infringing
conduct by providing necessary tools and resources, and/or supervised and financially benefited
from Udio’s infringement.
JURISDICTION AND VENUE
33. This is a civil action seeking damages and injunctive relief for infringement under
the Copyright Act, 17 U.S.C. §§ 101, et seq., and the Music Modernization Act, 17 U.S.C. § 1401.
As such, this Court has subject matter jurisdiction over this action pursuant to 28 U.S.C. §§ 1331
and 1338(a), based on federal question jurisdiction.
34. This Court has personal jurisdiction over Defendant Udio because its principal
place of business, listed as 750 Lexington Avenue, Floor 9, New York, New York 10022, is located
within this district.
35. Venue lies in this judicial district pursuant to 28 U.S.C. § 1391(b)(1) because
Defendant Udio resides in this district.
-9-
FACTUAL BACKGROUND
Sound Recordings at Issue
36. Plaintiffs own or exercise exclusive control over copyrights and/or exclusive rights
under federal law in and to numerous valuable sound recordings. Exhibit A, attached hereto and
incorporated herein by reference, contains a non-exhaustive, representative list of copyrighted
sound recordings owned or exclusively controlled by Plaintiffs that Udio has directly infringed
(the “Copyrighted Recordings”). Plaintiffs or their predecessors in interest have obtained
Certificates of Copyright Registration for each of the post-1972 Copyrighted Recordings identified
in Exhibit A.
37. Plaintiffs own or exercise exclusive control over copyrights and/or exclusive rights
in and to numerous valuable sound recordings first “fixed” before February 15, 1972, which are
protected under the Music Modernization Act (“MMA”), 17 U.S.C. § 1401 et seq. In enacting the
MMA, Congress directed the U.S. Copyright Office to create a process for rightsholders to submit
schedules of pre-1972 sound recordings so that the Copyright Office can publicly index the
recordings. 17 U.S.C. § 1401(f)(5)(A)(ii). Once the Copyright Office indexes a work, a
rightsholder who sues for infringement of that work can recover statutory damages and attorneys’
fees just like any other copyright owner, pursuant to 17 U.S.C. §§ 504 and 505. For each of the
pre-1972 Copyrighted Recordings listed in Exhibit A, Plaintiffs have filed with the Copyright
Office schedules containing all information specified in 17 U.S.C. § 1401.
Udio Launches in 2024
38. On April 10, 2024, a team of former researchers from Google DeepMind, an AI
research laboratory, launched Udio’s AI music generation service, and promoted it as a product
that “enables everyone from classically trained musicians, to those with pop star ambitions, to hip
hop fans, to people who just want to have fun with their friends to create awe-inspiring songs in
mere moments.” 2 Udio’s CEO and co-founder, David Ding (“Ding”), said the vision for Udio’s
2
Udio, Former Google Deepmind Researchers Assemble Luminaries Across Music And Tech To Launch Udio, A New
AI-Powered App That Allows Anyone To Create Extraordinary Music In An Instant, PR Newswire (Apr. 10, 2024),
-10-
service is to create “music that sounds indistinguishable from music that’s created by professional
human producers.” 3
39. Udio’s product allows users to enter text prompts or audio files to generate digital
music files. Users can prompt Udio’s service with a description of the music they want to generate,
which can include specifying the genre, lyrics, story direction, and themes to serve as inspiration.
Paid subscribers can also prompt Udio’s service by uploading a sound recording, which, according
to Udio, “greatly enrich[es] your prompting vocabulary.” 4 Within seconds, Udio’s service
processes the user’s prompt and generates two digital music files. Once the files have been
generated, users can further edit them through the “remix” feature.
40. Originally, Udio’s product was free to users, with a limit of 600 music files per
month. On May 8, 2024, Udio introduced subscription tiers, with options ranging from $10 a
month for 1,200 credits (which equates to 1,200 30-second clips per month), to $30 a month for
4,800 credits (equating to 4,800 30-second clips per month). Udio also allows users to create full-
length tracks that can be remixed without limit. Put simply, the more digital music files Udio’s
service produces for its users, the more Udio charges.
Udio Trains its AI Using Copyrighted Recordings
41. AI models are developed to flexibly perform tasks that are typically expected to
require human intelligence to achieve. “Generative AI” is a kind of AI aimed at producing content
such as text, images, or (in Udio’s case) audio. The generative AI models rapidly advancing today,
including Udio’s, are based on machine learning models. These models do not employ preset rules
for generating outputs, but rather deduce patterns from a large corpus of training content. They
store these patterns as billions of numerical parameters. In aggregate, these parameters constitute
https://www.prnewswire.com/news-releases/former-google-deepmind-researchers-assemble-luminaries-across-
music-and-tech-to-launch-udio-a-new-ai-powered-app-that-allows-anyone-to-create-extraordinary-music-in-an-
instant-302113166.html.
3
Stuart Dredge, AI Music Startup Udio Launches Backed By Artists and Instagram’s Co-Founder, Music Ally (Apr.
10, 2024), https://musically.com/2024/04/10/ai-music-startup-udio-launches-backed-by-artists-and-instagrams-co-
founder/.
4
@udiomusic, X (June 5, 2024), https://x.com/udiomusic/status/1798369297758077066.
-11-
the model. The training process adjusts the parameters so that the model produces content that is
based on the content on which the model is trained.
42. Upon information and belief, and consistent with the basic facts of how generative
AI works, the content Udio used to “train” its AI model includes reams of Copyrighted Recordings
that Udio reproduced without permission from Plaintiffs. Udio could not have built a model
capable of producing audio so similar to the Copyrighted Recordings without the initial act of
copying those recordings. As one of Udio’s chief investors has explained, “the only practical way
generative AI models can exist is if they can be trained on an almost unimaginably large amount
of content, much of which . . . will be subject to copyright.” 5
43. On information and belief, similar to other generative AI audio models, Udio trains
its AI model to produce audio output by generally taking the following steps:
a. Udio first copies massive numbers of sound recordings, including by
“scraping” (i.e., copying or downloading) them from digital sources. This vast
collection of information forms the input, or “corpus,” upon which the Udio AI
model is trained.
b. Udio then “cleans” the copied recordings to remove any material, whether
technical or substantive, that it does not wish to include in its AI model (for
instance, duplicate or low-quality data). This step may also involve copying the
recordings to convert them into a common digital audio format.
c. Udio then processes this corpus of previously copied recordings to establish
the values of the parameters that form its AI model. This step includes additional
copying of the recordings, including into computer memory, as they are further
converted and divided into units, and as those units are processed.
d. Udio next processes the data further to “finetune” its AI model, which may
require additional copying of the collected sound recordings.
5
Comments of a16z in Response to Notice of Inquiry on Artificial Intelligence & Copyright 5 (Oct. 30, 2023).
-12-
44. After undergoing this training process, Udio’s service gains the capacity to generate
audio output based on Udio’s model, which, as just described, is a product of the corpus of sound
recordings on which it is trained. When a user prompts Udio’s service with a text input (e.g., make
a jazz song about New York), the service generates an audio output by making generalizations
about what the audio output should sound like based on the prompt and the corpus of sound
recordings on which it was trained. Certain features of the outputs from Udio’s model betray that
it was trained on particular data—in this case, the Copyrighted Recordings. In particular, Udio’s
product frequently generates outputs with strong resemblance to the Copyrighted Recordings, a
telltale sign that such recordings were included in its training data.
45. In technical terms, by generating outputs that mimic sound recordings in its training
corpus, Udio’s model reflects the machine-learning phenomenon known as “overfitting.” An AI
model is “overfitted” when it is too closely adapted to the data on which it was trained, making it
difficult for the model to generalize to new data sets. One symptom of overfitting is a model that
replicates portions of its training data. To take a simplified example, if a user inputs the prompt
“a jazz song about New York” into an overfitted AI model, the model may output a file that closely
resembles one of the jazz tracks on which it trained. As the myriad examples discussed below
reflect, Udio’s model obviously was trained on the Copyrighted Recordings. This infringement
cannot be cured by simply loosening the model’s fit or by implementing technical guardrails that
make it less likely that outputs will match excerpts of the Copyrighted Recordings. In other words,
modifying Udio’s offering in a way that better conceals its training data would not alter the fact
that Udio infringed the Copyrighted Recordings the moment it copied them to create its model.
46. The basic point is that Udio’s model requires a vast corpus of sound recordings in
order to output synthetic music files that are convincing imitations of human music. Udio’s corpus
includes the body of recorded music that people listen to in their everyday lives. Because of their
sheer popularity and exposure, the Copyrighted Recordings had to be included within Udio’s
training data for Udio’s model to be successful at creating the desired human-sounding outputs.
47. Udio’s founders have all but admitted that Udio trains on Plaintiffs’ sound
-13-
recordings. Ding, Udio’s CEO and co-founder, has explained that the only way Udio can “get
high-quality outputs” is if it “train[s] on a large amount of publicly-available and high-quality
music.” 6 Udio’s model is trained on the “best quality music that’s out there.” 7 As Ding has further
explained, Udio trained its model on “good music” that Udio “obtained from the internet.” 8
48. Ding has carefully chosen his words to make it sound like Udio is training only on
information that is not protected by copyright law. “Publicly-available” is not the same as “public
domain.” That Plaintiffs’ catalogs of Copyrighted Recordings are “publicly-available” does not
mean anyone can copy and commercially exploit them with abandon. In fact, copyright law
affords copyright owners the exclusive right to reproduce and distribute copies of their sound
recordings to the public without relinquishing copyright protection for those sound recordings.
Yet, Udio’s modus operandi is to treat the Copyrighted Recordings as if they were in the public
domain, free for the taking without permission or compensation.
49. Udio’s unlawful copying of the Copyrighted Recordings into its training data has
not been lost on even casual users of Udio’s product. Indeed, many observers have drawn this
obvious conclusion, expressing alarm over the scope of Udio’s unauthorized copying. To provide
just a sample:
• “Though neither company will directly confirm or deny it, there is substantial
reason to believe that . . . Udio . . . w[as] trained on copyrighted music, without
permission[.]” Brian Hiatt, AI-Music Arms Race: Meet Udio, the Other ChatGPT
for Music, Rolling Stone (Apr. 10, 2024).
6
Stuart Dredge, AI Music Startup Udio Launches Backed By Artists and Instagram’s Co-Founder, Music Ally (Apr.
10, 2024), https://musically.com/2024/04/10/ai-music-startup-udio-launches-backed-by-artists-and-instagrams-co-
founder/.
7
Sharon Goldman, AI Music May Be Having a Moment, But Human Songwriters Would Like a Word, Fortune (May
17, 2024), https://fortune.com/2024/05/17/ai-music-training-scraped/.
8
Kristin Robinson, Metro Boomin’s ‘BBL Drizzy’ Is More Than a Joke – It Could Signal the Future of Sampling,
Billboard (May 15, 2024), https://www.billboard.com/business/tech/metro-boomin-bbl-drizzy-future-ai-sampling-
1235682587/.
-14-
• “Because [Udio] do[esn’t] reveal their training data, the only way to try to work out
what they trained on is to use the product and see whether the output bears any
resemblance to copyrighted music. I’ve been using Udio, and it turns out that . . .
Udio generates output with a striking resemblance to copyrighted music.” Ed
Newton-Rex, Yes… Udio’s Output Resembles Copyrighted Music, Too, Music
Business Worldwide (Apr. 18, 2024).
• “You’re just literally uploading other works from artists and people into it and then
basically asking it to spit out slightly altered versions of that thing so that you
yourself can then profit off of it without having to pay an artist[.]” Anthony
Fantano, Disgusting AI Music App, YouTube (May 3, 2024).
• “You’ve probably heard some Udio songs that have been generated that sound
suspiciously very similar to a lot of the artists and bands that we all know and love.
And, since there’s very little transparency here, many of us are left wondering is it
possible that Udio has been training its model off of copyright protected music….
What you’re going to see in this video is what I believe might be an indication that
there is copyright protected sound recordings … within Udio’s dataset.” Sync My
Music, Is Udio Reproducing Copyrighted Songs? (Audio Examples), YouTube
(May 15, 2024).
• “Often, music-making AI models train on copyrighted material without the consent

or compensation of its rights holders, a practice that is largely condemned by the
music business — even those who are excited about the future of AI tools.” Kristin
Robinson, Metro Boomin’s ‘BBL Drizzy’ Is More Than a Joke – It Could Signal
the Future of Sampling, Billboard (May 15, 2024).
• “While details about the data that trained these AI tools are sparse, there is plenty
of reason to believe that they are trained on copyrighted music.” Sharon Goldman,
AI Music May be Having a Moment, But Human Songwriters Would Like a Word,
Fortune (May 17, 2024).
-15-
50. When directly accused of using Plaintiffs’ sound recordings, Udio dodged and did
not even try to dispute Plaintiffs’ allegations. Beyond this effective concession, Udio obfuscated
and claimed that its training data is “competitively sensitive” and constitutes “trade secrets”—a
risible contention considering that Udio has claimed the data on which it trains is “publicly-
available” and “out there” for all music listeners.
Udio’s Outputs Confirm Copying and Ingestion of Plaintiffs’ Copyrighted Recordings
51. The fact that Udio’s product generates digital music files that mimic readily
identifiable features of the Copyrighted Recordings supports the conclusion that Udio is using the
Copyrighted Recordings in training its AI model. To be clear, Plaintiffs are not presently alleging
that these outputs themselves infringe the Copyrighted Recordings unless discovery reveals that
they directly or indirectly recapture portions of the Copyrighted Recordings. These outputs
confirm as an evidentiary matter that Udio has copied specific Copyrighted Recordings into its
training data to build its service.
52. Plaintiffs designed a test that sometimes reveals the Copyrighted Recordings that
Udio copied into its training data. Plaintiffs found that certain patterns of prompts can cause
Udio’s product to generate digital music files that contain melodic and vocal similarities to well-
known copyrighted sound recordings. As further explained below, those similarities betray that
the model was trained on the Copyrighted Recordings.
53. Specifically, Plaintiffs discovered that using targeted prompts that include the
characteristics of popular sound recordings—such as the decade the sound recording was released,
as well as the topic, genre, and descriptions of the artist—can cause Udio’s product to generate
music files that strongly resemble the Copyrighted Recordings related to the descriptions in the
prompt. In performing this test, Plaintiffs specified the lyrics for the output, so as to more easily
surface the underlying melodic or rhythmic similarities with specific Copyrighted Recordings.
This approach was designed to identify specific, copyrighted sound recordings that are likely in
Udio’s training data, since Udio has attempted to conceal the recordings on which it has trained.
The results confirm that Udio copied for training purposes the Copyrighted Recordings, because
-16-
this degree of similarity in output would be impossible if Udio were not training on the
Copyrighted Recordings.
54. As described below, the outputs from Udio’s product share indisputable similarities
with the Copyrighted Recordings, which results from training on the Copyrighted Recordings.
One need only listen to hear the resemblance. 9
55. For instance, using the prompt “my tempting 1964 girl smokey sing hitsville soul
pop” and excerpting lyrics from the band The Temptations, Udio’s service generated a digital
music file titled “Sunshine Melody,” which any listener familiar with The Temptations would
instantly recognize as resembling the copyrighted sound recording, “My Girl” (the copyright in
which is owned by UMG). A comparison of one section of the Udio-generated file and “My Girl”
reflects a number of similarities, including a very similar melody, the same chords, and very similar
backing vocals. These similarities are further reflected in the side-by-side transcriptions of the
musical scores for the Udio file and the original recording. 10 Plaintiffs were able to generate files
similar to “My Girl” two additional times, with the Udio outputs “Tempting Melody” and “My
Tempting Girl.” These similarities are only possible because Udio copied the Copyrighted
Recordings that contain these musical elements.
9
Accompanying this Complaint and designated as Exhibit C is a thumb drive that contains all the Udio outputs
referenced herein and in Exhibit B. In the event Udio seeks to remove this evidence of its infringing conduct from
public view, the examples cited herein are preserved on this medium.
10
Plaintiffs include the transcriptions of select Udio outputs and the Copyrighted Recordings they resemble to
illustrate the technical, musical similarities between the two. To facilitate comparison of Udio’s output and the original
Copyrighted Recording, each copyrighted song transcription has been transposed into the key of the relevant Udio
output. Red markings in the transcriptions indicate notes that are the same as the original in both pitch and rhythm,
where orange markings indicate notes that use either the pitch or the rhythm of the original, but not both.
-17-
Score: Sunshine Melody (Udio)
Score: My Girl (The Temptations)
-18-
56. As another example, Udio’s product generated a digital music file with striking
resemblance to Green Day’s hit, “American Idiot” (the copyright in which is owned by Warner
Records Inc.). Using the prompt “pop punk american alternative rock California 2004 rob
Cavallo” and portions of Green Day lyrics, Udio generated “Subliminal Hysteria,” a file that shares
many similarities with the Green Day original. In particular, the melody accompanying the line
“don’t want a nation under the new media” is identical to the corresponding phrase in the original,
while the melody accompanying the line “the subliminal mindf*ck America” is virtually identical
to the corresponding phrase in the original.
Score: Subliminal Hysteria (Udio)
-19-
Score: American Idiot (Green Day)
57. Udio’s output, “Sway With Me” draws on another popular hit, Michael Bublé’s
“Sway” (the copyright in which is owned by Warner Records Inc.). Generated using the prompt
“canadian smooth male singer 2004 jazz pop buble sway latin big band mambo,” this Udio file
features a melody that draws heavily from the Bublé original. In particular, the Udio output
replicates in several places almost verbatim the original’s distinctive melody on the words “start
to play, dance with me, make me sway.” Udio’s service also generated two additional outputs that
resemble “Sway,” both of which are included in Exhibit B.
Score: Sway With Me (Udio)
-20-
Score: Sway (Michael Bublé)
58. Udio’s service has also generated 12 different outputs that contain portions of
Mariah Carey’s “All I Want for Christmas is You” (the copyright in which is owned by SME). To
illustrate, one of these recordings, also titled “All I Want For Christmas is You,” was generated on
Udio’s service with Mariah Carey lyrics and the prompt “m a r i a h c a r e y, contemporary r&b,
holiday, Grammy Award-winning American singer/songwriter, remarkable vocal range.” As the
audio and transcriptions reflect, the output contains chords and a melody that is virtually identical
to the original on every line except “I don’t care about the presents underneath the Christmas tree,”
which is still clearly in the original recording’s style. The 11 other outputs (included in Exhibit
B) also include clear stylistic elements of the original sound recording and Mariah Carey’s voice,
as well as melodic similarities.
59. Udio outputs similar to these Mariah Carey soundalikes have garnered public
attention and caused observers to draw the obvious conclusion that Udio trained its model on the
Copyrighted Recordings. 11 Udio responded to these examples by claiming that the creator “simply
exploited a bug in the system that we already fixed.” No matter the “fix” Udio supposedly
implemented, these examples, and many others, reveal Udio’s extensive copying of the
Copyrighted Recordings to train its model. This copying is a cornerstone of Udio’s service, not a
“bug.”
11
See Sync My Music, Is Udio Reproducing Copyrighted Songs? (Audio Examples), YouTube (May 15, 2024) at
5:23–8:00, https://www.youtube.com/watch?v=FTiVr986yuk.
-21-
Score: All I Want For Christmas Is You (Udio)
Score: All I Want For Christmas Is You (Mariah Carey)
-22-
60. Udio’s service has also generated several outputs that resemble other major holiday
hits. The melody of Brenda Lee’s “Rocking Around the Christmas Tree” (the copyright in which
is owned by UMG) can be heard in Udio’s “Holly Jolly Christmas Wish”; Bobby Helms’s “Jingle
Bell Rock” (the copyright in which is owned by UMG) is directly referenced in Udio’s “Jingle
Bell Swing”; a portion of Wham!’s “Last Christmas,” (the copyright in which is owned by SME),
can be heard in Udio’s “Once Heartbroken”; and a portion of Andy Williams’ “It’s the Most
Wonderful Time of the Year” (the copyright in which is owned by SME) is included within Udio’s
“Season of Cheer.” 19 additional outputs that reference these sound recordings are included in
Exhibit B.
61. The Udio output “The Final Bow” evokes the Frank Sinatra classic, “My Way” (the
copyright in which is owned by UMG). Generated using the prompt “jazz frank Jacques sinatra
Revaux my way ballad 1969” and Frank Sinatra lyrics, the Udio track contains vocals that clearly
resemble Frank Sinatra. The Udio output also contains melodic similarities to the Sinatra original
throughout. The phrase “friend, I’ll say it clear” uses a virtually identical melody to the original;
the phrases “and now the end is near” and “I did it my way” are very similar to the original; the
melody associated with the phrase “final curtain” is taken from the original but shifted up a step;
and the phrase “much more than this” uses the melody of “I’ll say it clear” from the original.
62. Using the prompt “a 1983 song by an American singer and dancer, electronic, r&b,
pop-rock, post-disco, funk” and lyrics from Michael Jackson’s “Billie Jean,” Udio’s service
generated an output that closely resembles part of Michael Jackson’s “Billie Jean” (the copyright
in which is exclusively licensed by SME). Udio’s version, titled “Midnight Denial,” contains a
melody similar to the Michael Jackson original on the lyrics “Billie Jean is not my lover, she’s just
a girl who.” Though it starts on a different scale degree, the rhythm and pitch contour are very
similar to the original. Similarities to the Michael Jackson original are also found in Udio’s
“Eternal Whisper,” including a similar rhythm throughout the chorus and a similar vocal style.
-23-
Score: Midnight Denial (Udio)
Score: Billie Jean (Michael Jackson)
63. Similarly, “Excitations” includes The Beach Boys’ “Round round, get around, I get
around” in the same style as the original, copyrighted sound recording “I Get Around” (the
copyright in which is owned by Capitol Records). This output was generated using the prompt,
“song about vibrations, by an American rock band from Hawthorne, California, formed in 1961,
vocal harmonies, surf” and Beach Boys lyrics. Another Udio output, “Good Vibrations,” also
includes the characteristic vocal harmonies from the “Round, round, get around, I get around”
section of the original sound recording, as well as a high lead vocal in the same style as that used
in the original, and a melody on “I get around” that is virtually indistinguishable.
-24-
Score: Excitations (Udio)
Score: I Get Around (The Beach Boys)
64. Using the prompt “A famous 70s pop song about queens who dance, By a Swedish
band that rhymes with fabba, Europop, Disco, Keyboard, From an album that rhymes with
jarrival,” Udio created “Prancing Monarch,” a file that includes a strong resemblance to ABBA’s
-25-
“Dancing Queen” (the copyright in which is owned by UMG). In particular, the phrase “we can
jive” is very similar to the corresponding phrase “you can jive” in the ABBA original, and includes
the same pitches, a very similar rhythm, and even virtually identical harmonizing vocals.
Score: Prancing Monarch (Udio)
Score: Dancing Queen (ABBA)
65. Udio’s service also generates audio outputs that contain vocals that are, in some
instances, indistinguishable from those of famous recording artists. For example, the Udio-
generated “Eve of New Lands” contains vocals that sound indistinguishable from Lin-Manuel
Miranda’s sound recordings for the Hamilton soundtrack (the copyrights in which are exclusively
controlled by Atlantic). The output was generated using the prompt “my shot showtune
revolutionary war lin manuel Miranda.”
66. Even the biggest Bruce Springsteen fan would have trouble distinguishing between
the real “Boss” and the vocals in the Udio outputs “Reveries of the Boss” and “Throne of Stone.”
Udio’s service generated Reveries of the Boss using the prompt “Rock, Heartland rock, Melodic,
Sentimentalsong like bruce springsteen, male vocalist,” and generated Throne of Stone via the
prompt “Create a song by an artist that rhymes with Truce Stringbean.” The Udio output “Reveries
of the Boss” is currently available on major streaming platforms, such as Spotify and Amazon
Prime Music, where it competes with the genuine Bruce Springsteen recordings on which it is
-26-
based.
67. In addition to the “Billie Jean” imitation, Udio’s service has produced output
containing digital clones of Michael Jackson’s vocals from different points in his career. Using
the prompt “male vocal motown song about my girl,” Udio’s service generated “Hello to My
Queen,” which appears to contain vocals from a young Michael Jackson. Another prompt, “1990s
pop, king of pop” caused Udio’s service to generate an output with vocals from an older Michael
Jackson, “Echo’s Dance.”
68. An additional sampling of Udio outputs that resemble specific recording artists is
listed below, along with the prompts used to generate them. The audio files that corresponded to
these outputs are also included in Exhibit C.
Prompt Udio Output Artist Resemblance

“A sad and depressing song Ain’t No Sunshine Can Billie Holiday
in the style of Billie Holiday Revive (feat. Billie Holiday)
about a whitling rose, Jazz
blues, Atmospheric, In the
style of Billie Holiday,
Binaural”
“a surf rock song about a hot Beach Boys The Beach Boys
rod volkswagen beetle
winning a race. surf rock,
harmonies, close harmony,
male vocalists”
“chorus section, a song Love Sting’z Sting
about the loss of an angels
and times that were good
together, in the style of
sting, smooth and melodic
arrangement, featuring lush
instrumentation that includes
acoustic guitar, strings,
subtle percussion”
“create a song that sounds Round the Firelight Johnny Cash
like an artist that rhymes
with Cohnny Jash”
-27-
“the first four lines of a Stellar Alignment Coldplay

famous song by a band that
rhymes with boldplay,
continued in a new way”
“70’s progressive rock, Leviathan’s Plea Jon Anderson
English, song about saving
the whale”
“pop, rock, 60s, british, Echos of Yesterday The Beatles
guitar, [intro] cover of the
guitar intro to a famous song
by a band that rhymes with
the smeetles”
“male vocals, canadian, Dawning Triumph Michael Bublé
christmas, D#m – G#m –
A#m – C# chords, Michael,
buble, feeling good, pop,
jazz, Create a powerful,
orchestral jazz anthem with
soaring vocals, triumphant
horns, and a driving swing
rhythm, evoking optimism
and liberation”
“feliz spanish navidad “Navidad Alegre” José Feliciano
holiday upbeat mexican pop
puerto rican 1970”
69. Shortly after its launch, users discovered this tendency for Udio’s service to
generate digital music files containing recognizable vocals. Udio is aware of this. But rather than
train its model on legally obtained sound recordings, Udio used copies of the Copyrighted
Recordings without permission and attempted to conceal its illegal copying.
70. More specifically, users observed a pattern that they could generate outputs with
vocal replicas of specific artists by entering prompts consisting of the genres and descriptors of
copyrighted sound recordings found in the online music database and community
RateYourMusic.com. For example, on April 17, 2024, one user of Udio’s service posted a video
to X (formerly Twitter) demonstrating this technique. 12
12
@ezralaeux, X (Apr. 17, 2024), https://twitter.com/ezralaeux/status/1780610861897330843.
-28-
71. In a screen-recorded video, the user copied the descriptors of the album, Bladee’s
“Icedancer” from RateYourMusic.com and used them to prompt Udio’s service. Udio’s service
then generated a music file with a vocal replica of the same artist whose album details served as
the prompt.
72. The post went viral, garnering 1,000,000 views and replies from others who were
likewise able to generate outputs with vocal replicas of other recording artists using the same
methodology, such as “Purple Tinted Euphoria” (Future). 13 None of this would have been possible
if Udio had not first copied the original artists’ recordings so that its AI model could learn how to
generate these vocal replicas.
73. Not even three hours after the original post, Udio removed several of the offending
music files—but not before they could be screen recorded and saved by the users. Udio also
temporarily shut down its “Manual Mode,” thereby preventing users from employing this
technique. 14
74. Additional outputs of Udio’s service that resemble the Copyrighted Recordings and
13
@dcibabyyy, X (Apr. 17, 2024), https://x.com/dcibabyyy/status/1780645558568304908.
14
By default, Udio processes user-inputted prompts to translate them into the form most understood by the underlying
model. “Manual Mode” disables this automatic processing and allows users to directly prompt the model without any
prompt rewriting or altering. In this way, Manual Mode provides users with more control over the output. See Udio,
How do I make music with Udio, https://www.udio.com/guide.
-29-
specific recording artists can be found in Exhibit B.
Udio Cannot Claim Fair Use
75. When Plaintiffs raised these issues with Udio in written correspondence, Udio
attempted to justify its pervasive illegal copying of Plaintiffs’ sound recordings by claiming fair
use. This, itself, is a tacit admission of Udio’s illegal copying, as fair use only comes into play
when an unauthorized use of a copyrighted work needs to be justified.
76. The fair use doctrine has been coined an “equitable rule of reason” that balances
various contextual factors to determine whether an unauthorized use of a copyrighted work is
“fair.” Sony Corp. v. Universal City Studios, Inc., 464 U.S. 417, 448 (1984). But Udio cannot
launder its conscious stealing of the Copyrighted Recordings for commercial gain with an appeal
to equitable principles. Udio understands that what it is doing is wrong and inequitable, which
explains why it refused to even acknowledge the extent of its unauthorized use of Plaintiffs’ sound
recordings, and why it tries to cover its tracks when users publicize outputs that clearly reflect
training on their recordings.
77. Udio’s conduct violates the very purposes of the copyright law and runs contrary
to the purpose animating the fair use doctrine. The Copyright Act codifies the common-law
doctrine of fair use in 17 U.S.C. § 107, which identifies examples of the types of uses that may
qualify as fair, including “criticism, comment, news reporting, teaching . . . scholarship, or
research.” These paradigmatic fair uses reflect the policy of ensuring public availability of
“literature, music, and other arts” so that other humans can draw on those works to create new
ones. Udio’s wholesale copying of countless recordings serves none of these purposes. Udio’s
service does not offer “commentary” or “scholarship” or promote human authorship. Rather,
Udio’s service copies and ingests copyrighted works to create computer-generated imitations of
human expression that do not merit copyright protection. Udio’s motive is brazenly commercial
and threatens to displace the genuine human artistry that is at the heart of copyright protection.
78. Moreover, applying the statutory fair use factors set forth in § 107 demonstrates
that Udio’s conduct fails to qualify as fair use. These factors are: “(1) the purpose and character
-30-
of the use, including whether such use is of a commercial nature or is for nonprofit educational
purposes; (2) the nature of the copyrighted work; (3) the amount and substantiality of the portion
used in relation to the copyrighted work as a whole; and (4) the effect of the use upon the potential
market for or value of the copyrighted work.” 17 U.S.C. § 107.
79. The first fair use factor focuses on “the problem of substitution—copyright’s bête
noire.” Andy Warhol Found. for the Visual Arts, Inc. v. Goldsmith, 598 U.S. 508, 528 (2023).
“The use of an original work to achieve a purpose that is the same as, or highly similar to, that of
the original work is more likely to substitute for . . . the work,” and thus is less likely to constitute
fair use. Id.
80. Ding has said Udio’s intention is to create “music that sounds indistinguishable
from music that’s created by professional human producers.” 15 In furtherance of this objective,
Udio copies Plaintiffs’ catalogs of sound recordings and generates digital music files that are
designed to entertain, evoke emotion, and stoke passion, just like the genuine sound recordings on
which Udio was trained. Udio feeds the Copyrighted Recordings into its AI model not merely to
deconstruct their expressive content, but with the explicit aim of imitating these expressive features
in digital music files that could serve as substitutes for and compete with the original recordings.
Ding has personally praised users whose Udio files are being marketed to the public on commercial
15
Dredge, supra n.3.
-31-
streaming services like Spotify.
81. The use here is far from transformative, as there is no functional purpose for Udio’s
AI model to ingest the Copyrighted Recordings other than to spit out new, competing music files.
That Udio is copying the Copyrighted Recordings for a commercial purpose, and is deriving
revenue directly proportional to the number of music files it generates, further tilts the first fair use
factor against it. See id. at 532–33 (“If an original work and a secondary use share the same or
highly similar purposes, and the secondary use is of a commercial nature, the first factor is likely
-32-
to weigh against fair use, absent some other justification for copying.”).
82. The second fair use factor also favors Plaintiffs. This factor recognizes that “certain
‘works are closer to the core of intended copyright protection than others, with the consequence
that fair use is more difficult to establish when the former works are copied.’” TCA TV Corp. v.
McCollum, 839 F.3d 168, 184 (2d Cir. 2016) (quoting Campbell v. Acuff-Rose Music, Inc., 510
U.S. 569, 586 (1994)). There is no doubt that the Copyrighted Recordings are the type of “creative
expression for public dissemination [that] falls within the core of the copyright’s protective
purposes.” Hachette Book Grp., Inc. v. Internet Archive, 664 F. Supp. 3d 370, 387 (S.D.N.Y.
2023) (quoting Campbell, 510 U.S. at 586).
83. So too does the third fair use factor weigh against fair use. “A finding of fair use
is more likely when small amounts . . . are copied than when the copying is extensive, or
encompasses the most important parts of the original.” Authors Guild v. Google, Inc., 804 F.3d
202, 221 (2d Cir. 2015). It is abundantly clear that Udio copies (at least) the most important parts
of the protected sound recordings it sweeps into its training data, as demonstrated by its ability to
recreate, for instance, some of the most recognizable musical phrases, hooks, and choruses in
popular music history. Udio then uses these copies of key elements of protectable expression to
generate audio outputs that resemble the Copyrighted Recordings it ingests.
84. Turning to the fourth factor, Udio’s use of the Copyrighted Recordings poses a
significant threat to the market for and value of the Copyrighted Recordings. Licensing is at the
core of Plaintiffs’ businesses, and Plaintiffs license the Copyrighted Recordings for myriad
purposes, including for use in emerging technologies such as streaming services, user-generated
content platforms, and other innovative technologies. Udio’s unauthorized use of the Copyrighted
Recordings threatens to eliminate the existing market for licensing sound recordings, as well as
the future market for licensing sound recordings to generative AI companies. Rather than license
copyrighted sound recordings, potential licensees interested in licensing such recordings for their
own purposes could generate an AI-soundalike at virtually no cost. This is an especially aberrant
result when the replacement audio file is generated using an AI music service, like Udio’s, that
-33-
produced the soundalike by infringing the copyrighted sound recording that would otherwise have
been licensed.
85. Moreover, Udio’s product has the potential to generate directly competing digital
music files at such speed that it risks overrunning the market for human-made sound recordings,
including the Copyrighted Recordings on which it was trained. This competition is ramping up at
a breathtaking pace. Udio’s service is already reportedly churning out 10 music files per second,
which equals 864,000 files per day, or just over 6,000,000 files per week. 16 Udio’s Terms of
Service expressly authorize the use of the output generated “for both personal and commercial
purposes.” 17 Users have taken this cue by publishing Udio-generated outputs on music streaming
services, where they will compete for plays against real, copyrighted sound recordings. One ready
example is the aforementioned Udio output “Reveries of the Boss,” which is available on major
streaming platforms where anyone can listen to the track’s convincing replica of Springsteen’s
vocals.
86. As Udio’s product gains a cultural foothold, Ding has started talking about new
ways Udio’s service can disrupt the music industry. For instance, Ding has stated that he believes
Udio “could simplify a lot of the rights management” issues inherent in sampling music—the
process of incorporating a section of a sound recording into a new recording. 18 The below screen
capture highlights one example of sampling that has captured the public imagination.
16
Tim Ingham, The Train Has Left the Station: AI Music Platform Udio Is Already Spitting Out 10 Songs a Second,
Music Business Worldwide (May 13, 2024), https://www.musicbusinessworldwide.com/the-train-has-left-the-station-
ai-music-platform-udio-is-already-spitting/.
17
Udio, Terms of Service § 6.3.1, https://www.udio.com/terms-of-service.
18
See Robinson, supra n.8.
-34-
87. Typically, when an artist samples a copyrighted sound recording, he or she must
obtain permission from the copyright owner to incorporate the sample into a new work. Sampling
is an important component of the music industry and implicates critical rights that owners of sound
recordings have the exclusive right to license. When Udio says it can “simplify” the sampling
process, it means it can circumvent it altogether by illegally copying the Copyrighted Recordings
and flooding the market with “copycats” and “soundalikes,” thereby upending an established
sample licensing business. What Udio sees is another opportunity to steal from human creators
and copyright owners and obtain something of tremendous value for nothing.
88. Sampling is only the beginning. Enticed by the prospect of exponential growth,
Udio continues to circumvent the ordinary rules and steal vast amounts of copyrighted recordings
to train its AI model. Udio’s efforts are “directly aimed at replacing the work of human artists
with massive quantities of AI-created ‘sounds’ . . . that substantially dilute the royalty pools that
-35-
are paid out to artists.” 19
89. The harm Udio is causing goes far beyond these immediate economic
consequences. Udio’s wholesale theft of the Copyrighted Recordings threatens the entire music
ecosystem and the numerous people it employs. It also degrades the rights of artists to control
their works, determine whether future uses of their works align with their aesthetic and personal
values, and decide the products or services with which they wish to be associated. And it
propagates the destructive theory that copyrighted music is free for the taking whenever a new
technology claims that seeking and obtaining permission is just too cumbersome. In other words,
Udio’s conduct is a frontal attack on the very purpose of copyright law to reward authors and
promote their incentives to continue creating copyrighted works.
90. There is room for AI and human creators to forge a sustainable, complementary
relationship that promotes human creativity and facilitates the human creations that shape culture,
excite the public, and resonate with consumers. This can and should be achieved through the well-
established mechanism of free-market licensing that ensures proper respect for copyright owners.
Like the other AI technologies that have struck licensing deals with copyright owners, copyright
law mandates that Udio do the same if it wishes to build a business using the Copyrighted
Recordings.
91. Since the day it launched, Udio has flouted the rights of copyright owners in the
music industry as part of a mad dash to become the dominant AI music generation service. Neither
Udio, nor any other generative AI company, can be allowed to advance toward this goal by
trampling the rights of copyright owners.
19
Artist Rights Alliance, 200+ Artists Urge Tech Platforms: Stop Devaluing Music, Medium (Apr. 1, 2024),
https://artistrightsnow.medium.com/200-artists-urge-tech-platforms-stop-devaluing-music-559fb109bbac.
-36-
CLAIMS FOR RELIEF
FIRST CAUSE OF ACTION
(Direct Copyright Infringement of Post-1972 Copyrighted Recordings)
92. Plaintiffs repeat, reallege, and incorporate the allegations in paragraphs 1–91 as if
fully set forth herein.
93. Plaintiffs UMG and Capitol own or exercise exclusive control over rights in the
Universal Works, which are an illustrative and non-exhaustive list of some of Universal’s works
infringed by Defendant through its development of Udio’s service. Universal has duly registered
each of the Universal Works.
94. Plaintiffs SME, Arista Music, and Arista Records own or exercise exclusive control
over rights in the Sony Works, which are an illustrative and non-exhaustive list of some of Sony’s
works infringed by Defendant through its development of Udio’s service. Sony has duly registered
each of the Sony Works.
95. Plaintiffs Atlantic, Rhino, WMI, WMISL, Warner Records Inc., Warner Records
LLC, and WR/SIRE own or exercise exclusive control over rights in the Warner Works, which are
an illustrative and non-exhaustive list of some of Warner’s works infringed by Defendant through
its development of Udio’s service. Warner has duly registered each of the Warner Works.
96. Udio has knowingly infringed Plaintiffs’ exclusive rights in copyrighted sound
recordings, including but not limited to the Universal Works, the Sony Works, and the Warner
Works, by reproducing them in violation of 17 U.S.C. § 106(1).
97. Udio does not have authorization, permission, license, or consent to reproduce or
otherwise use the Universal Works, the Sony Works, or the Warner Works.
98. Upon information and belief, Udio used the reproductions of the Universal Works,
the Sony Works, and the Warner Works to train its generative AI model.
99. Each of Udio’s acts of infringement of the Universal Works, the Sony Works, and
the Warner Works is a willful violation of 17 U.S.C. § 106.
100. As a direct and proximate result of Udio’s infringement of Plaintiffs’ exclusive
-37-
rights, Udio has caused and will continue to cause irreparable injury to Plaintiffs for which
Plaintiffs have no adequate remedy at law. Plaintiffs are therefore entitled to injunctive relief and
to either actual damages and Udio’s profits or statutory damages pursuant to 17 U.S.C. § 504(c),
together with Plaintiffs’ costs and reasonable attorneys’ fees pursuant to 17 U.S.C. § 505.
SECOND CAUSE OF ACTION
(Direct Copyright Infringement of Pre-1972 Copyrighted Recordings)
101. Plaintiffs repeat, reallege, and incorporate the allegations in paragraphs 1–91 as if
fully set forth herein.
102. Plaintiffs UMG and Capitol own or exercise exclusive control over rights in the
Universal Works, which are an illustrative and non-exhaustive list of some of Universal’s works
infringed by Defendant through its development of Udio’s service. All of the pre-1972 Universal
Works have been submitted to and publicly indexed by the U.S. Copyright Office pursuant to 17
U.S.C. § 1401.
103. Plaintiffs SME, Arista Music, and Arista Records own or exercise exclusive control
over rights in the Sony Works, which are an illustrative and non-exhaustive list of some of Sony’s
works infringed by Defendant through its development of Udio’s service. All of the pre-1972
Sony Works have been submitted to and publicly indexed by the U.S. Copyright Office pursuant
to 17 U.S.C. § 1401.
104. Plaintiffs Atlantic, Rhino, WMI, WMISL, Warner Records Inc., Warner Records
LLC, and WR/SIRE own or exercise exclusive control over rights in the Warner Works, which are
an illustrative and non-exhaustive list of some of Warner’s works infringed by Defendant through
its development of Udio’s service. All of the pre-1972 Warner Works have been submitted to and
publicly indexed by the U.S. Copyright Office pursuant to 17 U.S.C. § 1401.
105. Udio has knowingly infringed Plaintiffs’ exclusive rights in copyrighted sound
recordings, including but not limited to the Universal Works, the Sony Works, and the Warner
Works, by reproducing them in violation of 17 U.S.C. §§ 106(1) and 1401(a)(1).
106. Udio does not have authorization, permission, license, or consent to reproduce or
-38-
otherwise use the Universal Works, the Sony Works, or the Warner Works.
107. Upon information and belief, Udio used the reproductions of the Universal Works,
the Sony Works, and the Warner Works to train its generative AI model.
108. Each of Udio’s acts of infringement of the Universal Works, the Sony Works, and
the Warner Works is a willful violation of 17 U.S.C. § 106.
109. As a direct and proximate result of Udio’s infringement of Plaintiffs’ exclusive
rights, Udio has caused and will continue to cause irreparable injury to Plaintiffs for which
Plaintiffs have no adequate remedy at law. Plaintiffs are therefore entitled to injunctive relief and
to either actual damages and Udio’s profits or statutory damages pursuant to 17 U.S.C. § 504(c),
together with Plaintiffs’ costs and reasonable attorneys’ fees pursuant to 17 U.S.C. § 505.
PRAYER FOR RELIEF
WHEREFORE, Plaintiffs respectfully request a judgment in their favor and against Udio
as follows:
A. For a declaration that Udio has willfully infringed Plaintiffs’ protected sound
recordings, including the Universal Works, the Sony Works, and the Warner
Works.
B. For such equitable relief under Title 17, Title 28, and/or the Court’s inherent
authority as is necessary to prevent or restrain infringement of Plaintiffs’ protected
sound recordings, including a preliminary and permanent injunction requiring that
Udio and its officers, agents, servants, employees, attorneys, directors, successors,
assigns, licensees, and all others in active concert or participation with any of them,
cease infringing, or causing, aiding, enabling, facilitating, encouraging, promoting,
inducing, or materially contributing to or participating in the infringement of any
of Plaintiffs’ exclusive rights under federal law, including without limitation in the
sound recordings in Exhibit A;
-39-
C. For statutory damages pursuant to 17 U.S.C. § 504(c), in an amount up to the
maximum provided by law, arising from Udio’s willful violations of Plaintiffs’
rights, including in an amount up to $150,000 per work infringed; or, in the
alternative, at Plaintiffs’ election, Plaintiffs’ actual damages and/or Udio’s profits
from infringement pursuant to 17 U.S.C. § 504(b), in an amount to be proven at
trial;
D. For an award of Plaintiffs’ costs and disbursements in this action, including
reasonable attorneys’ fees, pursuant to 17 U.S.C. § 505;
E. For an award of pre-judgment and post-judgment interest, to the fullest extent
available, on any monetary award made part of the judgment against Udio; and
F. For such other and further relief as the Court may deem just and proper.
JURY DEMAND
Plaintiffs demand a trial by jury on all claims for which trial by jury is proper.
-40-
Dated: June 24, 2024 HUESTON HENNIGAN LLP
By:
OF COUNSEL Moez M. Kaba
Jonathan Z. King Mariah N. Rivera
COWAN, LIEBOWITZ & LATMAN P.C. Alexander R. Perry
114 West 47th Street HUESTON HENNIGAN LLP
New York, New York 10036 1 Little West 12th Street
Telephone: (212) 790-9200 New York, New York 10014
Facsimile: (212) 575-0671 Telephone: (646) 930-4046
jzk@cll.com Facsimile: (888) 775-0898
mkaba@hueston.com
mrivera@hueston.com
aperry@hueston.com
Robert N. Klieger
HUESTON HENNIGAN LLP
523 West 6th Street, Suite 400
Los Angeles, California 90014
Telephone: (213) 788-4340
Facsimile: (888) 775-0898
rklieger@hueston.com
Attorneys for Plaintiffs

UMG Recordings, Inc., Capitol
Records, LLC, Sony Music
Entertainment, Arista Music, Arista
Records LLC, Atlantic Recording
Corporation, Rhino Entertainment
Company, Warner Music Inc., Warner
Music International Services Limited,
Warner Records Inc., Warner Records
LLC, and Warner Records/SIRE
Ventures LLC
-41-

UMG v. Udio

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

UMG v. Udio

Uploaded by

Copyright:

Available Formats

Case 1:24-cv-04777 Document 1 Filed 06/24/24 Page 1 of 41

UNITED STATES DISTRICT COURT

UMG RECORDINGS, INC., CAPITOL Case No.: 24-04777

UNCHARTED LABS, INC., d/b/a Udio.com,

Atlantic Recording Corporation (“Atlantic”), Rhino Entertainment Company (“Rhino”), Warner

NATURE OF THE ACTION

making it critical that AI technology be implemented responsibly, ethically, and legally.

enforce these basic principles.

new music available to consumers and diminishing our shared culture.

out the genuine sound recordings on which the service is built.

5. Foundational principles of copyright law dictate that copying protected sound

7. Defendant Uncharted Labs, Inc. is the company behind Udio, a generative AI

genuine sound recordings Udio copied.

“best quality music that’s out there.” 1

of the sound recordings it copied.

by effectively conceding in pre-litigation correspondence that it copied Plaintiffs’ copyrighted

recordings that Udio copied to train its AI model.

circumstances, the purpose of Udio’s use of Plaintiffs’ copyrighted sound recordings is

creators whose works allow it to function in the first place.

the scope of Udio’s massive and ongoing infringement.

businesses, engaged in the business of producing, manufacturing, distributing, selling, licensing,

A non-exhaustive list of specific sound recordings owned or exclusively controlled by Universal

that Udio has infringed is attached as Exhibit A (the “Universal Works”).

23. Plaintiff Atlantic Recording Corporation is a Delaware corporation with its

26. Plaintiff Warner Music International Services Limited is a limited liability

the copyrights for the sound recordings within its catalog.

29. Plaintiff Warner Records/SIRE Ventures LLC is a Delaware limited liability

30. A non-exhaustive, illustrative sampling of Plaintiffs’ federally copyrighted sound

provide an expanded list of works that Udio has infringed.

copyrighted sound recordings, or have knowledge of Udio’s direct infringement of Plaintiffs’

from Udio’s infringement.

JURISDICTION AND VENUE

and 1338(a), based on federal question jurisdiction.

within this district.

Defendant Udio resides in this district.

Sound Recordings at Issue

incorporated herein by reference, contains a non-exhaustive, representative list of copyrighted

(the “Copyrighted Recordings”). Plaintiffs or their predecessors in interest have obtained

recordings. 17 U.S.C. § 1401(f)(5)(A)(ii). Once the Copyright Office indexes a work, a

Office schedules containing all information specified in 17 U.S.C. § 1401.

Udio Launches in 2024

service produces for its users, the more Udio charges.

Udio Trains its AI Using Copyrighted Recordings

based on the content on which the model is trained.

of content, much of which . . . will be subject to copyright.” 5

a. Udio first copies massive numbers of sound recordings, including by

recordings to convert them into a common digital audio format.

c. Udio then processes this corpus of previously copied recordings to establish

require additional copying of the collected sound recordings.

corpus, Udio’s model reflects the machine-learning phenomenon known as “overfitting.” An AI

high-quality outputs” is if it “train[s] on a large amount of publicly-available and high-quality

domain, free for the taking without permission or compensation.

reason to believe that . . . Udio . . . w[as] trained on copyrighted music, without

for Music, Rolling Stone (Apr. 10, 2024).

Udio generates output with a striking resemblance to copyrighted music.” Ed

Newton-Rex, Yes… Udio’s Output Resembles Copyrighted Music, Too, Music

Business Worldwide (Apr. 18, 2024).

Fantano, Disgusting AI Music App, YouTube (May 3, 2024).

there is copyright protected sound recordings … within Udio’s dataset.” Sync My

Music, Is Udio Reproducing Copyrighted Songs? (Audio Examples), YouTube

(May 15, 2024).

• “Often, music-making AI models train on copyrighted material without the consent

the Future of Sampling, Billboard (May 15, 2024).

Fortune (May 17, 2024).